Search Singularity - PyData Singapore

Published on: Monday, 1 May 2017

Search Singularity: Build Visual Search Engine using TensorFlow in GCE for 0$

Speaker: Nachi Nachiappan


Have we achieved Search Singularity? Well, we’re still evolving and TensorFlow provides one avenue to achieve this by powering Visual Search similar to what numerous 3rd party services (like ViSenze or Tineye or IQnect) offer. This talk discusses the design, framework, code and image sources to build your own Visual Search Service quickly and host it in Google Cloud. This search engine is built using a python crawler to source ecommerce images, python flask framework for search engine interface and TensorFlow and KNN for similarity search. Further the talk shares concepts on how to achieve the next level of search i.e. ‘Search Singularity’ for ecommerce sites.


Nachi Nachiappan works for Playware – a ‘Gaming as a Service’ provider and hacks and codes on full stack web development and data science. After an extensive stint at Avanade, he ventured into a start-up to digitize paper receipts using mobile NFC and having failed fast, moved into consulting and currently is passionately involved in data science competitions and codes in R and Python and explores new ML libraries. One notably recent ML library he dabbles on is by FB called Prophet for time series analysis.

Event Page:

Produced by Engineers.SG

Help us caption & translate this video!